> Does such a solution exist now. To me this appears to be in entirety
of
> what should constitute a database cluster. Only the search needs to be
> done on all the servers simultaneously at the low level. Once you get
the
> results, the writing can be determined by the upper level logic (which
can
> even be in a scripting language). But the search across many servers
has
> to be done using proper threading, and the re-sorting also needs to be
> done fast.
Well the fastest way would be to write a libpq wrapper, personally I
would choose C++ for extreme performance. STL bring super fast sorting
to the table and will make dealing with ExecParams/ExecPrepared a little
bit easier. To make available from scripting languages you need to make
C wrappers for interface functions and build in a shared library.
You could use any of a number of high level scripting languages but
performance will not be as good. YMMV.
Antother interesting take on this problem would be to use dblink
contrib. module. Check that out and see if it can meet your needs.
Merlin
